ESP32 Cam Object Detection Coco SSD Transfer Learning

[Florian Maier]

Hi, guys,

i found the example of ChungYi Fu on Github where he uses the Tensorflow Obeject Detection API to detect things with the ESP32 cam. https://github.com/fustyles/Arduino/blob/master/ESP32-CAM_Tensorflow.js/ESP32-CAM_coco-ssd/ESP32-CAM_coco-ssd.ino 

I would like to add a few more classes to the 80 available. For example other animals, vehicles, toys and co. Do any of you know a good guide on how I can do this? I have read that transfer learning can be used to further refine the Coco model, but how does it work? 

I would also be interested to know whether the model recognition can also run locally? The ESP32 Cam has a SD card reader. Up to now the model has been run online.

Friendly greetings

[Chris Pietsch]

Hello Florian,

there is a cool project using the ESP32 Cam and running a TF Lite Model on the device itself: https://github.com/jomjol/AI-on-the-edge-device/wiki

It is a meter for counting the water consumption of a house / flat. It is quite nicely documented and you could reuse the scripts for training the model: https://github.com/jomjol/neural-network-digital-counter-readout/blob/master/Train_CNN_Digital-Readout-Small-v1.ipynb

Hope this helps a bit.

Cheers Chris

--
You received this message because you are subscribed to the Google Groups "TensorFlow.js Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tfjs+uns...@tensorflow.org.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/tfjs/a33679d1-5fba-44ec-b9ee-e503c5f567d9n%40tensorflow.org.